Scheduling of crude oil and product blending and distribution operations in a refinery

159 CHAPTER 5 RECIPE DETERMINATION AND SCHEDULING OF GASOLINE BLENDING AND DISTRIBUTION OPERATIONS …..... The overall refinery operations involve three main segments, namely crude oil st

BLENDING AND DISTRIBUTION OPERATIONS

IN A REFINERY

JIE LI

NATIONAL UNIVERSITY OF SINGAPORE

2009

BLENDING AND DISTRIBUTION OPERATIONS

IN A REFINERY

JIE LI

(M.Eng., Tianjin University)

A THESIS SUBMITTED FOR THE DEGREE OF PHD OF ENGINEERING DEPARTMENT OF CHEMICAL AND BIOMOLECULAR ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

2009

I am very much thankful to my supervisors Professor I A Karimi and Professor

Rajagopalan Srinivasan for their enthusiasm, constant encouragement, insight and

invaluable suggestions ， patience and understanding during my research at the

National University of Singapore Their recommendations and ideas have helped me

very much in completing this research project successfully I would like to express my

heartfelt thanks to Professor I A Karimi and Professor Rajagopalan Srinivasan for

their guidance on writing scientific papers including this thesis

Special thanks go to all my lab mates, Mr P C P Reddy, Dr Li Wenkai, Dr

Liu Yu, Mr Ganesh Balla, Mr Suresh Pitty, Mr Arul Sundaramoorthy, Mr B Mohan

Babu, Ms Huang Cheng, Mr Suresh Selvarasu, Dr Mukta Bansal, Ms Maryam

Zargarzadeh, Mr M M Faruque Hasan, Mr Naresh Susarla, Ms Sangeeta Balram

and Dr Hong-Choon Oh, for sharing their knowledge with me I also wish to thank all

my friends for their constant encouragement and appreciation

I express my sincere and deepest gratitude to my parents, my younger sister, and

my relatives (Mr Yeo Kok Hong, Mrs Wu Jian and Miss Wu Jinjin) in Singapore, for

their boundless love, encouragement and moral support

Finally, I would like to thank the National University of Singapore for providing

a research scholarship to make this research project possible

ACKNOWLEDGEMENTS ……….……… i

SUMMARY ……….……… … viii

NOMENCLATURE ……….……… x

LIST OF FIGURES ……… xx

LIST OF TABLES ……… ……… xxiii

CHAPTER 1 INTRODUCTION ………. 1

1.1 Refinery Operations ……… ………. 2

1.2 The Supply Chain Management of Refinery ……… 4

1.3 Need for Management in Refinery Industry ……… 6

1.4 Supply Chain Management of Petroleum Industry ……… 6

1.5 Research Objective ……….……… 8

1.6 Outline of the Thesis ……….……. 9

CHAPTER 2 LITERATURE REVIEW ……… 12

2.1 Planning in Refinery ………. 12

2.2 Scheduling in Refinery Operation ………15

2.2.1 Crude Oil Scheduling ………. 17

2.2.2 Scheduling of Intermediate Processing ………. 24

2.2.3 Scheduling of Product Blending and Distribution Operation ……. 26

2.2.4 Scheduling of Product Transportation ……… 29

2.4 Uncertainty in Refinery Operations ………. 33

2.4.1 Reactive Scheduling ……… 33

2.4.2 Predictive Scheduling ……… 37

2.5 Summary of Research Gaps ……… ………. 39

2.6 Research Focus ……… 41

2.7 Time Representation ………. 43

CHAPTER 3 IMPROVING the ROBUSTNESS AND EFFICIENCY OF CRUDE SCHEDULING ALGORITHMS ……….……… 48

3.1 Introduction ………. 48

3.2 Problem Statement ……… 53

3.3 Base Formulation ………. 56

3.4 Motivation ……… 58

3.5 Extensions of Reddy’s Model ………. 60

3.6 Improving Robustness & Efficiency ……… 64

3.6.1 Backtracking Strategy ……… 67

3.6.2 Variables for Integer Cuts ……… 69

3.6.3 Revised Reddy’s Algorithm ………. 72

3.6.4 Partial Relaxation Strategy ……… 74

3.6.5 Algorithm Evaluation ………. 76

3.7 Solution Quality ……… 86

3.8 Upper Bound on Profit ……… 92

3.9 NLP-Based Strategy ………. 97

3.9.1 Evaluation of RLA ………103

3.10 Summary ………. 105

CHAPTER 4 A DISCRETE TIME MODEL WITH DIFFERENT CRUDE BLENDING POLICIES FOR CRUDE OIL SCHEDULING ……… 107

4.1 Introduction ………. 107

4.2 Problem Definition……… … 108

4.3 Mathematical Formulation ……….……… 111

4.4 Solution Method……… 128

4.5 Case Studies ……….… 131

4.5.1 Example 1……… 133

4.5.2 Examples 2-4 ……… 140

4.5.3 Examples 5-22 ……… …. 159

4.6 Summary ………. 159

CHAPTER 5 RECIPE DETERMINATION AND SCHEDULING OF GASOLINE BLENDING AND DISTRIBUTION OPERATIONS …

……… 161

5.1 Introduction ……… 161

5.2 Problem Statement ……… 165

5.3 Single-Period MILP ……… 170

5.3.2 Order Delivery ………. 177

5.3.3 Inventory Balance ……… 179

5.3.4 Transitions in Blenders ………. 180

5.3.5 Objective Function ……… 180

5.4 Schedule Adjustment ……… 181

5.5 Multi-Period Formulation ……… 187

5.6 Example 1 ………. 189

5.7 Detailed Evaluation ………. 203

5.8 MINLP Formulation ……… 216

5.9 Summary ……… 220

CHAPTER 6 INTEGRATING BLENDING AND DISTRIBUTION OF GASOLINE USING UNIT SLOTS ……… 221

6.1 Introduction ……… 221

6.2 Problem Statement ……… 222

6.3 Motivation ………. 225

6.4 MILP Formulation ……… 226

6.4.1 Blending and Storage ……… 229

6.4.2 Run Lengths and Product Quality ……… 232

6.4.3 Order Delivery ……… 234

6.4.4 Slot Timings on Component Tanks ………. 236

6.4.5 Slot Timings on Product Tanks ……… 238

6.4.7 Scheduling Objective ……… 239

6.5 Multi-Period Extension ……… 239

6.6 Schedule Adjustment ……… 240

6.7 Examples 1-2 ………. 249

6.8 Numerical Evaluation ………. 254

6.9 MINLP Formulation ………. 258

6.10 Summary ………. 259

CHAPTER 7 REACTIVE AND ROBUST CRUDE SHCEDULING UNDER UNCERTAINTY………….……… 260

7.1 Introduction ……… 260

7.2 Problem Statement ……… 262

7.3 Basic Formulation and Algorithm ………. 262

7.4 Reactive Scheduling ……… 264

7.4.1 Example 1 ……… 268

7.4.2 Example 2 281

7.5 Robustness Definition and Evaluation ……… 288

7.6 Demand Uncertainty ……… 290

7.6.1 Example 3 ……… 298

7.7 Summary ……… 299

CHAPTER 8 CONCLUSIONS AND RECOMMENDATIONS … 300

8.1 Conclusions ……… 300

REFERENCES ……… 305

_

Ever-changing crude prices, deteriorating crude qualities, fluctuating demands for products, and growing environmental concerns are squeezing the profit margins of modern oil refineries like never before Optimal scheduling of various operations in a refinery offers significant potential for saving costs and increasing profits The overall refinery operations involve three main segments, namely crude oil storage and processing, intermediate processing, and product blending and distribution This thesis addresses the first and third important components: scheduling of crude oil, and product blending and distribution

First, a robust and efficient algorithm is developed to solve large, nonconvex, mixed integer nonlinear programming (MINLP) problems arising from crude blending during crude oil scheduling The proposed algorithm solves all tested industrial-scale examples up to 20-day scheduling horizon However, commercial solvers (DICOPT and BARON) and the existing algorithms in the literature fail to solve most of them Moreover, the proposed algorithm gives profit within 6% of a conservative upper

bound In addition, the practical utility of Reddy et al (AIChE Journal, 2004b, 50(6),

1177-1197)’s MINLP formulation is enhanced by adding appropriate linear blending

correlations for fifteen crude properties that are critical to crude distillation and downstream processing, and controlling changes in feed rates of crude distillation unit (CDU)

Second, although the algorithm developed in the first part is intended for a marine-access refinery, the algorithmic strategy is successfully extended to in-land refineries involving both storage and charging tanks A general discrete-time formulation for an in-land refinery is developed and several crude blending polices in

storage and charging tanks are addressed Four literature examples and eighteen other examples with varying structures, sizes, and complexities are used to illustrate the capability of the proposed formulation and algorithm The results show that the proposed algorithm is superior to those in the literature

Third, a general synchronous slot-based MINLP formulation is developed for an integrated treatment of recipe, specifications, blending, storage, and distribution Many real-life features such as multi-purpose tanks, parallel non-identical blenders, constant rates during blending runs, minimum run lengths, changeovers, linear property indices, piecewise constant profiles for blend component qualities and feed rates, etc are incorporated in the model Since commercial MINLP solvers are unsatisfactory for solving this complex MINLP, a novel and efficient procedure that solves successive MILPs (mixed integer linear programming) instead of an MINLP, and gives excellent solutions is proposed

Fourth, a general and efficient MINLP formulation using unit slots is developed for the above blending and distribution problem This formulation incorporates all realistic features of the model proposed above Furthermore, it relaxes an assumption

to ensure sufficient supplies of components through the entire scheduling horizon By solving fourteen examples, it shows that the proposed unit-slot based model obtains the same or better solutions than the process-slot model with fewer binary variables and less computational time

Finally, a novel approach is first developed for reactive scheduling of crude oil operation Then, a scenario-based MINLP model is developed to obtain robust schedule for demand uncertainty during crude oil scheduling The obtained schedule is more robust than the initial schedule

Chapter 3

Notation

Sets

IC Set of pairs (tank i, crude c) that i can hold c

IU Set of pairs (tank i, CDU u) that i can feed u

IIU Set of pairs (tank ii, CDU u) that ii can feed u

IF11 iuct Dynamic set defined from the value of slack variable u iuct−

IF12 i(i ′)uct Dynamic set defined from the value of slack variable u iuct−

IF21 iuct Dynamic set defined from the value of slack variable u iuct+

IF22 i(i ′)uct Dynamic set defined from the value of slack variable u iuct+

IE1 (ii)ukt Dynamic set defined from the values ofs ukt− and s ukt+

θ Specification index for property k of crude c

NZ The number of terms in the first summation

XP pt 1 if parcel p is connected for transfer during period t

XT it 1 if tank i is connected to receive crude during period t

Y iut 1 if tank i feeds CDU u during period t

Continuous Variables

FTU iut Total amount of crude from tank i to CDU u during period t

FCTU iuct The amount of crude c from tank i to CDU u during period t

V it Total amount of crude in tank i at the end of period t

VCT ict The amount of crude c in tank i at the end of period t

f ict The fraction of crude c in tank i at the end of period t

FU ut Total amount of crude fed to CDU u during period t

u Crude Distillation Units (CDUs)

I Storage tank set

J Charging tank set

U Crude Distillation Units (CDUs) set

K Key component set

JP Set of jetty parcels

SP Set of VLCC parcels

PT Set of pairs (parcel p, period t) such that p can connect to SBM line during t

PI Set of pairs (parcel p, storage tank i) such that i may receive crude from p

IJ Set of pairs (storage tank i, charging tank j) that j may receive crude from i

JU Set of pairs (parcel p, CDU u) such that j can feed crude to CDU u

IC Set of pairs (storage tank i, crude type c) such that I can hold c

JC Set of pairs (charging tank j, crude type c) such that j can hold c

PC Set of pairs (pair p, crude type c) such that p is the last parcel of v

xkb Limits on the composition of key component k in charging tank i

xcp cp The composition of crude c in parcel p

D u Total crude demand per CDU u in the scheduling horizon

D ut Crude demand per CDU u in each period t

CP cu Margin ($/unit volume) for crude c in CDU u

COC Cost (k$) per changeover

SSP Safety stock penalty ($ per unit volume below desired safety stock)

SS Desired safety stock (kbbl) of crude inventory in any period

SWC v Demurrage or Sea waiting cost ($ per period)

ETD v Expected time of departure of vessel v

ETU p Earliest possible unloading period for parcel p

PS p Size of the parcel p

NJ Number of identical Jetties

Binary Variables

XP pt 1 if parcel p is connected to SBM/jetty discharge line during period t

XT it 1 if a tank i is connected to SBM/jetty discharge line during period t

Y iut 1 if a tank i feeds CDU u during period t

0-1 Continuous Variables

XF pt 1 if a parcel p first connects to the SBM/jetty during period t

XL pt 1 if a parcel p disconnects from the SBM/jetty during period t

X pit 1 if a parcel p and tank i both connect to the SBM line at t

YY iut 1 if a tank i is connected to CDU u during both period t and (t+1)

CO ut 1 if a CDU u has a changeover during period t

XSB ijt 0 if storage tank i does not feed charging tank j

Continuous Variables

TF p Time at which parcel p first connects to SBM/jetty for unloading

TL p Time at which parcel p disconnects from SBM/jetty after unloading

FPT pit Amount of crude transferred from parcel p to storage tank i during period t

FSB ijt Amount of crude transferred from storage tank i to charging tank j

FB jt Total amount of crude fed to charging tank j during period t

FCSB ijct Amount of crude c delivered by tank i to tank j during period t

FTU iut Amount of crude that tank i feeds to CDU u during period t

FU ut Total amount of crude fed to CDU u during period t

FCTU iuct Amount of crude c delivered by tank i to CDU u during period t

VCST ict Amount of crude c in storage tank i at the end of period t

VST it Crude level in storage tank i at the end of period t

VCBT jct Amount of crude c in charging tank j at the end of period t

VBT jt Crude level in charging tank j at the end of period t

fs ict Composition (volume fraction) of crude c in tank i at the end of period t

fb jct Composition (volume fraction) of crude c in tank j at the end of period t

DC v Demurrage cost for vessel v

SC t Safety stock penalty for period t

BJ Set of pairs (blender b, product tank j) such that blender b can feed product

tank j

PJ Set of pairs (product p, product tank j) such that tank j can hold product p

JO Set of pair (product tank j, order o) such that tank j can deliver order o

OP Set of pair (order o, product p) such that order o is for product p

TK Set of pair (slot k, period t) such that slot k is in period t

RL Minimum run length of blender b for product p

RL b Maximum of the minimum blend run lengths for blender b

M b Most volume that blender b can process during a slot

θ Limits on the index for property s of product p

θis Blend index of property s of component i

r Limits on the fraction of component i in product p

DR jo Delivery rate of product tank j to order o

DD Due date of order o

F i Constant feed rate of component i into its tank

/

L U

i

V Limits on the holdup in component tank i

c i Price ($ per unit volume) of component i

CB b Cost ($ per occurrence) of a transition on blender b

CT j Cost ($ per occurrence) of a transition in product tank j

DM o Demurrage ($ per unit time) for order o

R bk Rate of blender b in slot k

TCQ bk Volume processed by blender b during slot k

F it Constant feed rate of component i to its tank during period t

CRL bk Corrected run length of blender b at the end of slot k

CCQ bk Corrected volume processed by blender b during slot k

TCQ bk Total volume processed by blender b during the current run at the end of slot

k

θist Blend index for a property s of component i during period t

ρit Density of component i during period t

max

i

ρ Maximum possible density among all products during period t

N The number of products that are needed to process in blenders

N b The number of blenders

Binary Variables

v bjk 1, if blender b feeds product tank j (0 < j ≤ J) during slot k

u jpk 1, if product p is stored in product tank p during slot k

z jok 1, if product tank j is delivering order o during slot k

0-1 Continuous Variables

v b0k 1, if blender b is idle during slot k

ue jk 1, if product tank j switches products at the end of slot k

x bpk 1, if blender b produces product p during slot k

xe bk 1, if blender b ends its current run for a product during slot k

Continuous Variables

T k Time at which slot k ends

SL k Length of slot k

G bjk Volume that blender b feeds product tank j during slot k

VP jk Inventory in product tank j at the end of slot k

RL bk Length of the current run of blender b at the end of slot k

Q bk Volume processed in blender b during slot k

q ibk Volume of component i used by blender b during slot k

CQ bk Volume processed by blender b during the current run, if the run does not

end during slot k

DQ jok Volume of order o delivered by product tank j during slot k

d o Demurrage ($) for order o

V ik Inventory in component tank i at the end of slot k

TC Total operating cost ($)

F bk Rate of blender b during slot k

NP The number of distinct products that must be processed by blenders during

the scheduling horizon

B The number of blenders

T qk Time at which slot k on unit q ends

T ik Time at which slot k on component tank i ends

T bk Time at which slot k on blender b ends

T jk Time at which slot k on product tank j ends

BL bk Time for which blender b processes real products (p > 0) in slot k

ts jok the time at which the delivery of order o by product tank j begins in slot k

t ik An intermediate point between T i(k−1) and T ik

VC ik Inventory in component tank i at the end of slot k

PPXP pt 1, if the parcel-to-SBM/jetty connection changes

PPXT it 1, if the SBM/jetty-to-tank connection changes

PPY iut 1, if the tank-to-CDU connection changes

Figure 1.1 A simplified configuration of the petroleum industry ….…….………… 3 Figure 1.2 Schematic of a typical petrochemical supply chain ……….……… 5 Figure 1.3 A configuration of managerial activities in a refinery ……….…….7 Figure 2.1 Schematic of the overall refinery operation ………. 16 Figure 2.2 Classification of continuous-time scheduling models ……… 44 Figure 3.1 Schematic of crude oil unloading, blending, and processing ……….… 54 Figure 3.2 Flow chart for RRA [Revised Algorithm of Reddy et al (2004a,b)] …… 75 Figure 3.3 Schematic of RRA-P (Partial Relaxation Strategy) ……….… 76 Figure 3.4 Flow chart for RRA-P1 (Partial Relaxation Refinement Strategy) …… 91 Figure 3.5 Definition of sets for slack cuts ……….……….… 102 Figure 3.6 Flow chart for RLA [Revised Algorithm of Li et al (2002)] …………. 104 Figure 4.1 Schematic of crude oil unloading, storage, blending, and processing … 109 Figure 4.2 Flow chart for RRA-P1 (Partial Relaxation Refinement Strategy) ….… 132 Figure 4.3 Oil flow network for Example 1 ……… …….…… 133 Figure 4.4 Oil flow network for Example 2 ……… 140 Figure 4.5 Oil flow network for Example 3 ……….…… ………… 141 Figure 4.6 Oil flow network for Example 4 ……… 141 Figure 5.1 Schematic of gasoline blending and distribution ………….………… 164 Figure 5.2 Schematic of slot design ……….……… 169

Figure 5.3 An example schedule to illustrate intermittent delivery of orders O1 and O2

by PT-101 and PT-120 respectively ……… 184

Figure 5.4 The schedule of Figure 3 revised by our algorithm where 101 and

PT-102 deliver O1 and O2 continuously ……….…… 185

Figure 5.5 Flowchart for the schedule adjustment procedure ……… 186

Figure 5.6 Optimal schedule for Example 1 (5 orders) from SPM ……….………199

Figure 5.7 Optimal schedule for Example 1 (5 orders) from RSPM …….………. 200

Figure 5.8 Feed rate profiles of blend components from component tanks for Example

Figure 5.9 Optimal schedule for Example 1 (5 orders) from RMPM ………….… 202

Figure 5.10 Optimal schedule for Example 4 (15 orders) from RSPM ………… 209

Figure 5.11 Optimal schedule for Example 5 (15 orders) from RMPM …… ……210

Figure 5.12a Delivery schedule for Example 9 (23 orders) from RSPM ……….…211

Figure 5.12b Blending schedule for Example 9 (23 orders) from RSPM ……… 212

Figure 5.13a Delivery schedule for Example 12 (35 orders) from RSPM with

intermittent delivery of O13 by PT-109 ………. 213

Figure 5.13b The delivery schedule of Figure 13a revised by our algorithm where

PT-109 delivers O13 continuously ……… ……… 214

Figure 5.13c Blending schedule for Example 12 (35 orders) from RSPM ……… 215

Figure 6.1 A schedule using process slots ………….……… 227

Figure 6.2 The schedule using unit slots for Figure 6.2 ……….………… 228

Figure 6.3 Schematic of unit slots design ……….……… 229

Figure 6.4 An example for inventory violation of a component tank ….………… 237

Figure 6.5 Flowchart for the schedule adjustment procedure ……….………… 244

Figure 6.6 Optimal schedule for Example 1 (5 orders) from RSPM ……….……. 245

Figure 6.7 Optimal schedule for Example 2 (10 orders) from RSPM ……….……246

Figure 6.8 Optimal schedule for Example 1 (5 orders) from RMPM ……….…… 250

Figure 6.9 Optimal schedule for Example 2 (10 orders) from RMPM ….……… 251

Figure 6.10a Blending schedule for Example 12 (35 orders) from RSPM ….……. 252

Figure 6.10b Order delivery schedule for Example 12 (35 orders) from RSPM … 253

Table 3.8 Transfer rates, processing limits, operating costs, crude margins, and demands

Table 3.9a Specific gravities, sulfur contents, nitrogen contents, carbon residues for

crudes and acceptable ranges for feeds to CDUs ……… 83 Table 3.9b Pour points, freeze points, flash points, smoke points, Ni contents and Reid

vapor pressures for crudes and acceptable ranges for feeds to CDUs ……84 Table 3.9c Asphaltenes, aromatics, paraffins, naphthenes, and viscosities for crudes and

acceptable ranges for feeds to CDUs ……… 85 Table 3.10 Solution statistics for various algorithms/codes ……… 87 Table 3.11 Operation schedule from RRA-P1 for Example 16 ……… 89 Table 3.12 The upper bound for Examples 1-21 ……… 99 Table 4.1 Constraints for different refinery configurations and crude blending

Table 4.2 Data for Example 1 ……… 134 Table 4.3 Model and solution statistics for Example 1 with different cases ………. 135 Table 4.4 Data for Example 2 ……… 136

Table 4.5 Data for Example 3 ……… 137

Table 4.6 Data for Example 4 ……… 138

Table 4.7 Computational performance for Examples 2-4 ……… 139

Table 4.8 Proposed operation schedule for Example 4 ………. 142

Table 4.9 Ship arrival data for Examples 5-22 ……… 143

Table 4.10 Tank capacities, heels, and initial inventories for Examples 5-22 …… 144

Table 4.11a Initial compositions of crudes C1, C2, C5, and C6 for Examples

Table 4.11b Initial compositions of crudes C3, C4, C7, and C8 for Examples

Table 4.12 Crude concentration ranges in tanks and CDUs for Examples 5-20 …… 147

Table 4.13 Transfer rates, processing limits, operating costs, crude margins, and

demands for Examples 5-22 ……… 148

Table 4.14a Specific gravities, sulfur contents, nitrogen contents, carbon residues, pour

point, freeze point, and flash point for crudes and acceptable ranges for

Table 4.14b Smoke point, Ni, Reid vapor pressure, asphaltenes, aromatics, paraffins,

naphthenes, viscosity for crudes and acceptable ranges for feeds to

Table 4.15 Different operation features for Examples 5-22 ……… 151

Table 4.16 Model and solution statistics for Examples 5-22 ……… 152

Table 4.17a Operation schedule for vessel unload, storage tank receipt and feed for

Table 4.17b Operation schedule for charging tank feed to CDU for Example 7 ……154

Table 4.18a Operation schedule for vessel unload, and storage tank receipt and feed for

Table 4.18b Operation schedule for charging tank feed to CDU for Example 16 … 156

Table 4.19a Operation schedule for vessel unload, and storage tank receipt and feed for

…… ………

Table 4.19b Operation schedule for charging tank feed to CDU for Example 22 … 158

Table 5.1 Gasoline properties, corresponding indices, and correlations ………… 176

Table 5.2a Order data for Examples 1-9 ……… 191

Table 5.2b Order data for Examples 10-14 ………. 192

Table 5.3 Product and component tank data for Examples 1-14 ………. 194

Table 5.4 Component and product property indices for Examples 1-14 ………… 195

Table 5.5 Allowable composition ranges for components in products of Examples

Table 5.6 Blender and economic data for Examples 1-14 ……….…… 197

Table 5.7 Periods, slots, and feed flow rates to component tanks for Examples

Table 5.8 Computational performance of SPM ……… 206

Table 5.9 Computational performance of MPM ……… 207

Table 5.10 RMIPs and best possible solutions for Examples 1-14 from SPM …… 208

Table 5.11 Solution statistics of various algorithms/codes for SPM for Examples

Table 7.1 Data for Example 1 ……… 270

Table 7.2 The initial schedule for Example 1 (Profit = $ 1849K) ……… ……… 271

Table 7.3 Proposed schedule with RDM for Example 1−Parcel 7 delayed to the end of

period 7, informed at the end of period 4 (Profit = $ 1848.73K) ………. 272

Table 7.4 Proposed schedule with RRDM for Example 1−parcel 7 delayed to the end of

period 7, informed at the end of period 4 (Profit = $ 1838.58K) ….… 274

Table 7.5 Proposed schedule with RRDM for Example 1−Tank 4 unavailable from

period 2-4, informed at the end of period 1 (Profit = $ 1834.26K) … … 275

Table 7.6 Proposed schedule with RRDM for Example 1−CDU 3 demand increases

from 400 to 450, informed at the end of period 4 (Profit = $

Table 7.7 Proposed schedule with RRDM for Example 1−VLCC delayed by 3 periods,

informed at the end of period 1 (Profit = $ 1930.80K) ……… 277

Table 7.8 Proposed schedule with RRDM for Example 1−Tank 2 unavailable in periods

4-6, and concurrently the demand of CDU 2 increases from 400 kbbl to 440

kbbl, informed at the end of period 2 (Profit = $ 1895.08K) … …… 278

Table 7.9 Proposed approach vs block preservation for Example 1 ……… 279

Table 7.10 An alternative initial schedule for Example 1 ………. 280

Table 7.11 Proposed schedule with RRDM for Example 2−parcel 6 delayed two

periods, informed at the end of period 4 (Profit = $

Table 7.12 Proposed schedule with RRDM for Example 2−Tank 2 unavailable from

periods 2 to 3, informed at the end of period 1 (Profit = $

Table 7.13 Proposed schedule with RRDM for Example 2−CDU 1 demand increases

from 400 to 450, informed at the end of period 4 (Profit = $

Table 7.14 Proposed schedule with RRDM for Example 2−SBM pipeline unavailable

from periods 2 to 5, informed at the end of period 1 (Profit = $

1831.96K) …… 285

Table 7.15 Proposed schedule with RRDM for Example 2−Tank 3 unavailable from

periods 4 to 5 and demand of CDU 2 decreases from 400 to 387.5

simultaneously, informed at the end of period

2 ……….…… 286

Table 7.16 Proposed approach vs block preservation for Example 2 ………. 287

Table 7.17 Data for Example 3 ……… 294

Table 7.18 An initial schedule for Example 3 ……… …………. 295

Table 7.19 Proposed robust schedule for Example 3 ……… ……… 296

Table 7.20 Computational result for Example 3 ………. 297

CHAPTER 1

INTRODUCTION

During the last century, the petroleum industry has risen from being relatively small to

a position where whole economies are profoundly influenced by the need for and prices of petroleum products The petroleum business involves many independent operations, beginning with the exploration for oil and gas and extending to the delivery

of finished products, with complex refining processes in the middle These processes turn crudes into a wide range of products including gasoline, diesel, heating oil, residual fuel, coke, lubricants, asphalt, and waxes Unlike batch manufacturing industry such as food and pharmaceutical industries, petroleum refinery is typically a continuous process plant that has a continuous flow of materials going in and coming out In recent years, globalization has made the refining industry an extremely competitive business characterized by fluctuating demands for products, ever-changing raw material prices, and incessant push towards cleaner fuels Facing these stringent situations, refineries seek efficient managerial tools and apply new technology to maximize profit margins and minimize wastes simultaneously to improve their operations The following sections briefly introduce refinery operations, the entire supply chain of petroleum industry, its managerial activities, etc

1.1 Refinery Operations

Crude oil as the basic raw material of the petroleum industry is explored at different fields that are located in different countries all over the world such as Brazil and Middle East, and transported from these fields to refineries by vessels, trains, or oil pipelines for refining After its arrival, crude is stored or mixed in tanks, then charged

to the crude distillation unit (CDU) and is separated into several component streams (distillation cuts) such as light gases, propane, butanes, light naphtha, heavy naphtha, kerosene, light gas oil, heavy gas oil, vacuum gas oils and residue, whose boiling points lie within certain ranges e.g 30℃-130℃, 130℃-270℃, 270℃-370℃, etc Some of these streams are desirable, while others are undesirable The undesirable fractions are either sent to the downstream units for further treatment and undergo specific unit operations and processes in separate units such as Fluid Catalytic Cracker (FCC) unit, hydrocracking (crackers), hydrotreating, reformers, alkylation and isomerization units to yield desirable products by chemically altering the hydrocarbon molecules, splitting them or removing sulfur for instance The desirable products have

a wide range of physical properties such as density, viscosity, sulfur content, pour point, flash point, Reid vapor pressure (RVP), vanadium and nickel content On their own, these desirable products may not be suitable for commercial use, but when blended together or with those desirable streams in various ways, they form final products, which are known as Liquefied Petroleum Gas (LPG), gasoline, diesel, kerosene, etc These final products are stored in the corresponding product tanks and then delivered

to the customers by trains, trucks, pipelines or ships

Figure 1.1 A simplified configuration of the petroleum industry (Http://www.energy.ca.gov/oil/refinery_flow.html)

Additionally, undesirable streams may be sold off or used as low-cost fuels Figure 1.1 shows a general configuration of the petroleum industry The entire industry involves crude storing and mixing tanks, crude distillation units (CDU), vacuum distillations units (VDU), catalytic reforming units, fluid catalytic/hydro cracking units, hydro treaters, visbraker/delayed coker units and off-site storage/blending facilities to store/process the finished products/intermediate streams

1.2 The Supply Chain of Refinery

Figure 1.2 shows a schematic of a typical petrochemical supply chain (Srinivasan et al 2006) Crude oil is first produced from either ground fields or offshore platforms After pretreatment and storage, it is transported via supertankers like VLCCs (Very Large Crude Carriers) and small vessels such as single-parcel vessels to various refineries around the world, unloaded through SBM (Single Buoy Mooring), SPM (Single Point Mooring), or jetty pipelines, stored and blended in storage or charging tanks, or both, and charged to CDU for processing It is then converted into a variety of intermediate bulk chemicals that are used as feeds to the petrochemical plants globally and consumer products such as fuels that are used in aviation, ground transport, electricity generation, etc Thus, a refinery supply chain involves three manufacturing centers, namely the oil fields & platforms, and the petroleum refineries, that are surrounded by

a host of logistics services in the forms of storage, transportation, distribution, packaging, etc (Srinivasan et al 2006)

1.3 Need for Management in Petroleum Industry

In the past, the petroleum industry has succeeded by creating markets and supplying them with suitable products Today, globalization has become an irreversible trend with the rapid development of Information Technology and the decreasing costs of communication and transportation Furthermore, new market dynamics such as the proliferation of product grade specifications, the drive for lower inventories, increased capital investments, environmental regulations, refinery retail and transportation asset rationalization, and higher market volatility are all adding to the complexity To survive financially, refineries have to seek efficient managerial tools and apply new technologies

In the refining processes, one key challenge is how to best operate the plant under different feed compositions, production rates, energy availability, ambient conditions, fuel heating values, feed and product prices, and many more factors that are changing all the time Undesirable changes may lead to off-spec products, reduced throughputs, increased equipment wear and tear, uncertainty and more work Past experience can achieve operating targets in some situations However, in order to fully exploit the complete spectrum of how the plant can be operated to maximize operating profit, efficient management using advanced computer-aided techniques is also needed

in the competitive environment

1.4 Supply Chain Management of Petroleum Industry

The main managerial activities of a refinery can be divided into three layers: planning,

scheduling and unit operations Optimization plays an important role in managing the oil refinery Oil refineries have used optimization techniques for a long time, specifically Linear Programs (LPs) for the planning and scheduling of process operations Planning and scheduling primarily differ in terms of the time frames involved Planning is generally undertaken for longer time horizons such as months or years and includes management objectives, policies, etc besides immediate processing requirements It represents aggregated objectives and usually does not include finer details The main objective of planning is to maximize the gross refinery profit margin while meeting demand forecast and efficiently using facility resources such as plant capacities, utilities, and manpower Optimal plan produced in the planning stage forms the basis for scheduling While scheduling defines the detailed specification of each unit at each time over a short horizon ranging from shifts to weeks to satisfy the targets set at the planning stage, the objective of scheduling is implementation of the plan subject to the variability that occurs in the real world This variability can be in feed stock supplies, quality, production process, customer requirements or transportation

Figure 1.3 A configuration of managerial activities in a refinery (Li, 2004)

First of all, the plant-wide plans issued by the head office who considers plant-wide factors such as market condition, raw material availability, operation capacity and so

on are sent to the scheduling office as guidelines These plans mainly handle business decisions such as which units to operate, which raw materials to process and which products to produce, etc The objective of planning is to obtain an optimal operation strategy that can maximize the total profit After analyzing the plan, the scheduling office determines detailed operation schedule for each unit that is to be executed in a plant within the scheduling horizon The objective of scheduling is to seek a feasible operation strategy that meets the planning requirements while maximizing the total profit These feasible schedules are sent to the unit operation office as the operation guidelines so that the operators can control the unit operations rigorously to realize the scheduling objective

With an effective supply chain management, the refinery can reduce costs of purchased crude oils and chemicals, feedstock, their quality issues, optimize and manage crudes and product inventory, increase plant yields, improve visibility of scheduling and inventories across the supply chain, and satisfy customers, etc

1.5 Research Objectives

This research focuses on refinery planning and scheduling operations While refinery-planning problems have been extensively studied and are considered well developed, as discussed in the next Chapter, scheduling problems can involve enormous considerations for conceiving an optimum schedule taking into account all

the factors In the most general form, the problem is too complicated to formulate mathematically, let alone solving and obtaining an optimum schedule And even if the problem is formulated, a simplistic approach of enumeration of alternatives sounds preposterous because of the number of possibilities that might exist (combinatorial nature of the problem) A lot of research has been undertaken in this area in the past decade with a focus on the development of exact and approximate methods to solve short-term scheduling problems Therefore, this research project focuses on real, large-scale scheduling problems during refinery operations Furthermore, some disruptions may be unavoidable during the refinery operations The focus of this research is also to take into account these disruptions, while developing optimal schedules to make them robust and efficient

With this, the objectives of this research work are to (1) Develop efficient mathematical models for scheduling refinery operations such as crude oil operations, and product blending and distribution operations, which incorporate many real operation features; (2) Develop new robust and efficient algorithms, for instance, decomposition algorithm to solve the developed models, especially for real large-size industrial problems; (3) Define and evaluate robustness and Develop robust schedules for refinery scheduling operations in the presence of uncertainties

1.6 Outline of the Thesis

This thesis includes eight chapters After a brief introduction in Chapter 1, Chapter 2 presents a detailed literature review on planning and scheduling of refinery operations

Based on this detailed review, several gaps in the existing work are summarized

In Chapter 3, the first part of scheduling operations in a refinery: crude oil scheduling is presented in detail Some deficiencies of the existing work in the literature are overcome Some strategies and ways are developed to improve robustness, quality, and solution speed of the algorithm in the literature, and estimate solution quality by means of a tight upper bounding strategy Twenty-four large simulated examples are used to demonstrate numerically the robustness and effectiveness of the improved algorithm In addition, the most important nonlinear crude properties that are crucial to crude distillation and downstream processing are identified and incorporated into the problem formulation

Chapter 4 extends the enhanced formulation and developed solution strategies in Chapter 3 to handle crude oil scheduling in an in-land refinery, in which crudes are stored or stored and blended in storage tanks and blended in charging tanks Three policies of crude concentrations are analyzed in storage and charging tanks: 1) constant crude composition in storage tanks but variable in charging tanks, 2) variable crude composition in both storage and charging tanks, 3) variable crude composition in storage tanks but prefixed in charging tanks

In Chapter 5, scheduling of gasoline blending and distribution operations is addressed A global slot-based continuous-time formulation for simultaneous treatment

of recipe, blending, scheduling, and distribution is developed A schedule adjustment procedure is proposed to solve the nonlinearity arising from ensuring constant blending rates of blenders during bend runs In addition, nine nonlinear important

properties for gasoline such as octane number, Reid vapor pressure (RVP), sulfur, benzene, and aromatics content are also accounted for

In Chapter 6, a novel unit-slot based continuous time formulation is developed to treat the same problem presented in Chapter 5 The novel formulation incorporates all real-life operation features of the model developed in Chapter 5 The basic formulation

is extended to multi-period scenario

Chapter 7 first uses reactive approach to address several disruptions during scheduling of crude oil operations and compare with the heuristic method of Arief et al (2007a) Then, schedule robustness is defined as schedule effectiveness, predictability and rescheduling stability Based on this, schedule robustness index (RI) is defined A procedure is proposed to evaluate the robustness of a schedule A scenario-based formulation is developed to obtain robust schedules for demand uncertainty

Finally, conclusions and recommendations for future research are summarized in Chapter 8